Previous studies have demonstrated the utility of creating multimeric targeting proteins for the purposes of either augmenting effective affinity through the formation of multivalent molecules or broadening the spectrum of recognition through the formation of multiple specific molecules. A variety of protein interaction domains have been employed to generate recombinant proteins with dimeric and multimeric binding sites. Initially, fusions of the targeting domains to leucine zipper domains were commonly used for dimerization. In this approach, hydrophobic interaction of leucine zipper domains is mediated by regularly spaced leucines in parallel α-helices, while the dimerization partner is determined by other amino acids immediately outside of the hydrophobic core, mainly charged residues, forming salt bridges (1-3). This interaction is exemplified by the Fos and Jun family of proteins, which preferentially form heterodimers without significant interference of the target domain specificity. This approach provides a versatile scaffold to create multimeric complexes (4,5). However, there are limitations that significantly affect the usefulness of this approach for therapeutic protein development. Most prominently, Fos and Jun are intracellular proteins that accumulate almost exclusively within the nucleus. Thus, soluble and secreted Fos and Jun fusions are usually produced using the baculovirus-infected or stably transformed insect cell system, a relatively low yielding and not easily scalable manufacturing process (6,7). In an attempt to create bispecific molecules, antibody domains linked to Fos-Jun were produced in bacterial or mammalian cells, but the main limitation was subunit homodimerization (8,9) which complicated the purification process and reduced the overall yield (8,9). Furthermore, the difference in patterns of glycosylation of proteins produced by insect or bacterial cells raises concerns of potential immunogenicity of the products when used in therapeutic applications.
In addition to leucine zipper motifis, immunoglobulin (IgG) constant domains, helix-turn-helix self dimerizing peptides, tri- and tetrameric subdomains of collagen and p53 have been used as scaffolds by which to create multivalent molecules (8, 10-13). Aside from the IgG fragments, these interaction domains primarily serve as molecular scaffolds and lack other functional activities per se. Moreover, fusion proteins containing these domains often require further optimization to promote stable multimer formation and specialized production cell lines and purification methods that are tedious or impose regulatory hurdles for therapeutic development (10,12). Many of these scaffolds are derivatives of either nonhuman protein domains or non-native components of plasma that may exhibit poor pharmacokinetic properties and pose the risk of immunogenic responses that could limit their therapeutic potential.